Power injection becomes stochastic due to large-scale wind power integration into power grid, which greatly challenges the traditional power grid. If the uncertainty is modelled using the interval number, the solution of economic dispatch model can also be represented as an interval number. To achieve the precise upper and lower bound, two deterministic programming models were deduced: the optimistic solution can be obtained by an optimistic model using linear programming, whereas the pessimistic solution achieved by a pessimistic model is an NP-hard problem, which can be transformed into a bilinear programming by duality theory. Then a linear relaxation technology, spatial branch and bound method were used to obtain the global pessimistic solution. Therefore, the interval bound information can provide the operator with institutive bound information for the future security assessment. Finally, a test system with 15 generators and 300 nodes was studied at different time scales. The result, compared with interior point method and enumeration, shows the effectiveness of proposed method.